Machine learning techniques to build geometrical transformations for object matching a review

Abstract

Image matching or object matching is one of the cutting edge research fields in machine learning or computer vision domain. Whereas aim of image matching techniques is to build geometrical transformations over source image and target image, videos, real time moving object to extract similarity measure. Several research methods devised for image matching but efficiency of techniques is bounded with various parameters such as image rotation, speed, blurriness, quality etc., these parameters are important while understanding and devising robust image matching techniques. Study and analysis of image matching parameters is highly important while learning and understanding, predicting performance when time is a limiting factor for implementation. Several approaches have been presented to achieve efficiency over real time object matching. Now in this paper we have presented fundamentals of object matching based on geometrical transformation to match object. Comprehensive review of existing methods with analysis of image matching parameters is presented to determine the limitations of existing methods. This review also addresses comparative study of existing image matching techniques to generalize criteria for design of robust technique.